CI-PASM - Computational Intelligence Based Prognostic Automotive System Model

In an ideal case physically oriented vehicle models can reduce the required practical knowledge of a vehicle designer. These types of models are effective cost reducing tools used in industrial development cycles. There are many variables that can be used as input both internal and external to model automobile performance. The focus of this paper is on those external variable factors such as environment conditions that are not controllable by a human but are instantaneously measurable and affect performance. This paper presents CI-PASM, A Computational Intelligence Based Prognostic Automotive System Model. Initial feature reduction was accomplished by a human expert. Principal Component Analysis was performed to further reduce the input set. Using expert chosen features, the CI-PASM algorithm produced results having an error at worst in the hundredths of a second. These output results were compared against a support vector machine implementation and were shown to be superior. The CI-PASM mean error was half that of the support vector machine error. Results from using PCA attributes and a support vector machine indicated that these are relevant alternative methods given different requirements.